The Silent Threat: How AI Chip Tariffs Could Trigger a Global Slowdown in Innovation

Until we master neuromorphic processors and asynchronous communication, technologies that replicate the processing morphology and energy efficiency of the human brain, GPUs remain our best option for powering traditional neural network models, including large language models (LLMs). Many believe one path toward Artificial General Intelligence (AGI) lies in evolving LLMs, enhancing their cognitive capabilities by reducing hallucinations and improving reasoning through brain-inspired computational models.

Now, beyond triggering major economic disruption and raising the risk of a global recession, the recently imposed U.S. tariffs on semiconductor imports may unintentionally slow down the progress of AI innovation. Ironically, a policy that could well be aimed at restricting China’s rise in AI might end up empowering it instead, potentially accelerating its path to becoming the dominant global tech superpower.

The Global AI Chip Supply Chain: An International Dance

To understand the magnitude of the disruption, it is useful to trace the end-to-end lifecycle of a modern GPU, which is foundational to AI model training and deployment. This journey involves many countries, each playing a critical role:

1. Chip Design: Companies like NVIDIA and AMD design GPU architectures in the U.S. using Electronic Design Automation (EDA) tools from Synopsys and Cadence.
2. Raw Materials and Silicon Production: The silicon used in chips is derived from quartz, with China and the U.S. being key producers. This silicon is then refined into metallurgical-grade polysilicon. (China dominates the global polysilicon market, controlling over 75% of production)
3. Wafer Production: The polysilicon is melted into ingots and sliced into wafers (or in other words the silicon is refined into wafers) in Japan, South Korea, and Germany.
4. Chip Fabrication : The actual chip is imprinted onto wafers by foundries like TSMC (Taiwan) and Samsung (South Korea), using lithography machines primarily from ASML (Netherlands). (TSMC manufactures roughly 50% of the world's semiconductors)
5. Testing and Packaging: Chips are cut and packaged by OSAT (Outsourced Semiconductor Assembly and Test) providers in Malaysia, Thailand, and China.
6. Chip Integration: The GPUs are integrated into boards in China or Vietnam.
7. Commercialization: These are then shipped globally and installed in data centres by companies like Amazon, Google, Microsoft, and Meta.

How the Tariffs Are Shaking the Semiconductor Supply Chain

On April 5, 2025, the U.S. imposed a 10% baseline tariff on all semiconductor imports, with steeper tariffs of 54% on Chinese goods, 32% on Taiwanese, and 46% on Vietnamese electronics. These countries form the backbone of the AI chip supply chain.

• China exports over $34 billion in AI-enabled servers to the U.S. annually and now faces the harshest tariffs.
• Taiwan, home to TSMC, the main chip manufacturer for NVIDIA and AMD, is burdened with a 32% tariff.
• Vietnam, an emerging hub for networking and board assembly, also faces a major hit.

Key companies are already feeling the effects:

• NVIDIA saw its stock drop over 7% post-announcement. Although some production is shifting to Arizona, domestic scaling is years away.
• Intel, while more U.S.-based, still depends on rare earths largely sourced from China.
• TSMC shares also fell by more than 7%, reflecting investor concern about rising costs and potential disruption.

The Paradox: Slowing Progress or Empowering China?

Tariffs can limit China’s access to advanced chips and design tools. It is true that China does not yet manufacture high-performance GPUs like NVIDIA’s H100, but they are experts in reverse engineering and rapidly closing the gap.

China has already retaliated with counter-tariffs and may restrict exports of rare earth elements, critical for chip manufacturing. While they currently rely on U.S. and Taiwanese GPUs, these restrictions could catalyze deeper domestic investment into alternatives.

As an example, China remains the only nation to have independently proven to launch and operate its own space station, using entirely their own launch vehicles, astronauts, and spacecraft. If any country can eventually build its own lithography machines and design competitive high-end GPUs, it’s China.

Rather than contain China's AI ambitions, the tariffs may ignite them as these short-term setbacks may lead to long-term gains through public investment and tech nationalism. As for the US and for the rest of the world, AI development may slow due to increased costs and unstable supply chains. Cloud providers may delay launching cutting-edge models and nations around the world may diversify or relocate chip production to regions like Southeast Asia, Mexico, or Europe, potentially leading to supply fragmentation and cost inflation.

Conclusion

The U.S. tariffs risk creating the very conditions they hoped they could prevent: a more self-reliant, innovation-driven China dominating the future of AI and semiconductors. Instead of halting China’s rise, we may be accelerating a global tech divide triggering in a new cold war, slowing collective AI advancement, and risking long-term instability.

In my view, international collaboration is the key to sustainable technological evolution. While these geopolitical tensions might incidentally delay the arrival of AGI and perhaps spare us from a SkyNet-like scenario, I believe humanity is capable of achieving far more through cooperation than confrontation.

Before the wave of tariffs and supply chain fragmentation, we witnessed unprecedented advancements in AI, enabled not necessarily by active international collaboration, but at least by a relatively undisturbed global semiconductor ecosystem. The absence of heavy trade barriers allowed innovation to flourish. Just imagine what we could achieve if that environment continued: Level-5 autonomous driving, intelligent home robots that shop for us and safely escort our children to school, fully autonomous robotic missions exploring every corner of our solar system, you name it.

This is the kind of future we should aim for, a future shaped not by artificial divisions, but by intelligent cooperation.

Disclaimer: The views expressed in this article are my own and do not represent those of ESA